FUNDING EMERGING ECOSYSTEMS
Documents de travail GREDEG GREDEG Working Papers Series
Paige ClaytonMaryann FeldmanBenjamin Montmartin
GREDEG WP No. 2019-25https://ideas.repec.org/s/gre/wpaper.html
Les opinions exprimées dans la série des Documents de travail GREDEG sont celles des auteurs et ne reflèlent pas nécessairement celles de l’institution. Les documents n’ont pas été soumis à un rapport formel et sont donc inclus dans cette série pour obtenir des commentaires et encourager la discussion. Les droits sur les documents appartiennent aux auteurs.
The views expressed in the GREDEG Working Paper Series are those of the author(s) and do not necessarily reflect those of the institution. The Working Papers have not undergone formal review and approval. Such papers are included in this series to elicit feedback and to encourage debate. Copyright belongs to the author(s).
Funding Emerging Ecosystems
Paige Clayton*, Maryann Feldman*, and Benjamin Montmartin**
University of North Carolina at Chapel Hill*
SKEMA Business School, Université Côte d'Azur**
GREDEG Working Paper No. 2019-25 September 2019
Abstract: Although prior research argues that location is important for firm performance, we lack an understanding of how resources accumulate in regions and how innovative ecosystems emerge and evolve over time. This paper focuses on the temporal development of an industry in a region and provides a framework for characterizing phase changes in a geographically defined entrepreneurial ecosystem. We add to the literature on entrepreneurial ecosystems by considering emergence as a temporal process and explicating finance as a mechanism that transitions between phases. Emergence is captured by the accumulation of individual entry by entrepreneurs, and punctuated by phase changes as the region accumulates resources and evolves. We use threshold regression to identify inflection points in stages of industry emergence. We then focus on the role of finance, from both public and private sources. Using a dynamic random effects probit model with regime analysis, we demonstrate interrelationships between public and private funding sources that differ over time. Finally, we estimate the relationship between the funding from various sources and firm survival within different phases using a discrete event history analysis. Our results demonstrate that public and private funding sources are complementary but with different impacts on firm survival during different phases. The results have implications for startup firms seeking funding and for policy making trying to encourage industry emergence. Acknowledgements: Funding for the development of the PLACE database was provided by the National Science Foundation and the Kauffman Foundation. We thank Makha Mooen for Bioscan data and Christos Kolympirisa for NIH data. We would also like to thank Mercedes Delgado and participants at AOM, APPAM, and Danish Research Unit for Industrial Dynamics (DRUID).
1
INTRODUCTION
Our understanding of the emergence of Silicon Valley, the archetype of a geographically
defined entrepreneurial ecosystem, has evolved from attributing primacy to venture capital to
recognizing the importance of early seeding from federal grants (Baker 2017; Lécuyer 2006;
Kenney & Von Berg 2000). Understanding how industries take hold and transform regional
economies is critical for both public and private investment and also for entrepreneurs who find
they are attracted to agglomerations. Our understanding of the spatial-temporal geographic
emergence of new industries coincides with other exploration of the dynamics of industry
emergence (Moeen & Agarwal 2017). Current economic development theory and policy advice
ignores the intricate historical processes inherent in the creation of an industry in a regional
economic context. Until scholars provide more evidence on how funding affects regional dynamics,
investors and policymakers will continue to make decisions with little evidence to guide them.
In considering emergence within a local entrepreneurial ecosystem, the focus has been on
the entry and survival of start-up firms, and the availability of resources (Spigel & Harrison 2017).
We heed Stam’s (2015) call for more theory development on entrepreneurial ecosystems. Through
an empirical assessment of historical context, we gain numerous insights from a careful study of
industrial development within its regional ecosystem. While social networks and support
organizations certainly play a role in industrial ecosystem development, the literature on new
industry financing is generally underdeveloped (Goldfarb et al. 2012). Influencing financial
investment can be useful to the myriad places trying to create innovative local economies.
In order to explain geographic ecosystem emergence, we analyze the long term
development of an industry in one region. Specifically, we examine the development of the life
sciences industry in the Research Triangle Region from 1980 to 2016. We focus on entrepreneurial
entry and trace the interacting roles of state and federal public funding and private funding of new
firms. Specifically, we ask how the interplay of these three sources of funding influenced the
emergence of the Research Triangle region’s life science industry. Emergence is straightforward to
track in this example because there was very little life science activity that existed previously,
either in the region or nationally. Of paramount importance to entrepreneurial ecosystem
emergence is the need for new ventures to survive long enough to contribute as a part of the whole.
We find that the steady hand of government provides financing to allow entrepreneurs to realize
opportunity. Over time, the private sector, attracted by these opportunities, becomes active as an
investment vehicle. Thus, an industry emerges through public-private interaction. Key to the idea of
2
emergence, though, is that these interactions are nonlinear with phase changes moving the system
along.
This paper makes the following contributions. First, we add more theoretical and analytic
rigor to the study of regional entrepreneurial ecosystems. We introduce that is the culmination of a
long term project that aims to take a long term perspective on ecosystem development. Our second
contribution is methodological: to address phase changes in the ecosystem, we provide a method
for detecting ecosystem life-cycle transition points. Finally, we contribute to our understanding of
how finance from blended sources affects ecosystem development.
THEORETICAL BACKGROUND
In examining the industrial evolution of a regional economy, the product life cycle model, a
biological metaphor, is often used. It is modeled by the canonical S-curve, where growth is some
function of time against a metric of industry dynamics, such as entry, with three or four defined
stages. The application of the life cycle model to geographic clusters has been criticized as
atheoretical and lacking a mechanism for an evolutionary framing (Bathelt & Boggs 2003;
Essletzbichler & Rigby 2007). Moving to an ecosystem framework, and away from a consideration
of clusters, permits a more dynamic focus on the context in which entrepreneurs, as the prime
actors in industry evolution, operate and an opportunity for appreciative theorizing.
In reviewing industry life cycle models, Agarwal et al. (2002) find theoretical explanations
from evolutionary economics, technology management, and organizational ecology. Each stream
offers differing arguments for the movement between phases, emphasizing the process of shakeout
as the industry moves from growth to maturity: evolutionary economics attributing maturity to
routinization; technology studies to the appearance of dominant designs; and organizational
ecology to population density. These theories are still dominant (Wang et al. 2014). These theories
may be applicable to understanding why certain regions become dominant in an industry but do
not address the question of how an industry initially begins in a specific location.
Recently, attention has turned to industry emergence (Agarwal et al. 2017; Moeen &
Agarwal 2017; Stine et al. 2015). Prior frameworks did not explain the initial birth phase of an
industry and the conditions that precede growth and industry shakeouts. The concept of emergence
is especially important for the development of a geographically defined ecosystem. While the
importance of geography to novel ideas and innovation is well established, we lack an
understanding of the dynamics of this advantage. As regions emerge, there is consensus that a
3
variety of endogenous forces are at work but these are not well specified (Martin & Sunley 2011).
Our understanding of the process of geographic emergence can only be understood within the
context of temporal dynamics. Organizational ecology offers one potential avenue yet our view of
emergence is based on the actions of agents which encompass more than population dynamics.
Finance and Ecosystem Emergence
Of course, building ecosystems requires financial resources. Publicly funded R&D is
associated with increased local firm births (Chen & Marchioni 2008; Kolympirisa et al. 2014; Zucker
et al. 1998). Kolympirisa et al. (2014) empirically demonstrate that public funds to firms are
associated more strongly with local firm births when compared with funds directed toward
universities, research institutes and hospitals, establishing the importance of examining firm level
R&D awards in order to understand geographic dynamics.
Private sources of new firm funding, once entrepreneurs move beyond friends and family,
include venture capital (VC) and angel investors (see Drover et al. 2017 for a review). A common
question in policy discourse is the effectiveness of government R&D subsidies relative to private
funds. Skeptics and opponents argue that public funds may crowd-out private funds while
proponents argue that public funding alleviates the market failures associated with the private
sector investment below a socially optimal level (David & Hall 2000). Montmartin & Massard
(2015) further argue that public funding creates knowledge externalities that are important to
locations with young firms. Samila & Sorenson (2010) find that public research funding and VC are
complementary and result in more innovative activity, as measured by patents and startups, when a
region has a greater amount of VC. Yet, public R&D financing programs, especially for high-risk, new
technologies are provided by a multitude of federal, state and local agencies, involving questions of
multi-level governance and coordination between different funding sources (Charbit 2011; Durufle
et al. 2018). Scholars have looked at public and private funding, or multi-level public funding, but
are rarely able to obtain data to examine all of these sources together.
Related work uses university research as the unit of analysis. Blume-Kohout et al. (2015)
are able to distinguish federal funding from non-federal sources, which include philanthropic
organizations, industry, and state government. Using the National Institute of Health (NIH) budget
doubling in late 1990s and early 2000s as an exogenous shock, they find that federal research
funding at universities crowded-in funding from these non-federal sources for biomedical research.
Using more recent data, Lanahan et al. (2015) find an increase of 1% increase in federal funding, “is
associated with a 0.411% increase in nonprofit research funding, a 0.217% increase in state and
4
local research funding, and a 0.468% increase in industry research funding,” arguing these results
indicate that federal government induces complementary investments by other sources. With
regard to federal versus local funding, Wu & Merriman (2017) find a substitution effect between
federal and state funding of university research, with effect depending on federal R&D spending
trends. A stronger effect is found during times of slow growth in federal R&D spending. Wu &
Merriman suggest that local funds may be used to stabilize the amount of financial resources
available.
Research has considered the interrelation between private and federal R&D funding for
firms (Zúniga-Vicente et al. 2014). For example, Lerner (1996) finds that federal SBIR funding
serves as a certification effect to entice private funding for entrepreneurial firms. David et al.
(2000) find disagreement about the degree to which public funding is a complement or substitute
for private funding, noting that the literature is mostly based on cross-sectional or limited time
series data, with limited nuance about the firm’s context. In a 10-year panel study of 316 MSAs, Lee
(2018) found no significant regional employment or income growth effects from SBA guaranteed
small business loans. In contrast, Minniti & Venturini (2017) find R&D tax credits have a significant
and economically important effect on the rate of long-term productivity growth. Durufle et al.
(2018) analyze emerging ecosystems around universities and suggest that further research
consider the appropriate combination of different types of policies for emergence.
EMPIRICAL CONTEXT
North Carolina's Research Triangle region's life science cluster ranks 5th in the US for life
science clusters, following Boston-Cambridge, the Bay Area, San Diego, and New Jersey (CBRE
Research 2019). What makes the Research Triangle unique is that it has been able to establish
preeminence without any industrial strength in related sectors before the 1980s. Unlike Cambridge
and the San Francisco Bay area—the country's leading life science regions—the Research Triangle
region was not an obvious candidate to develop a life science industry (Markusen 1996; Luger &
Goldstein 1991). The origins of the Research Triangle region’s life science cluster can be traced
back to the establishment of Research Triangle Park in 1958. The Park was the result of a
collaborative effort involving politicians, academics, and financiers to make the region more
competitive (Leyden & Link 2011). Over time, in the region adjacent to and surrounding the Park,
an ecosystem developed.
5
-------------------------
Insert Figure 1 about here
-------------------------
Figure 1 displays an increasing number of active life science entrepreneurial firms founded
in the Triangle over the time period analyzed. The ecosystem is defined as dedicated biotechnology
firms, human therapeutics, diagnostics, medical devices, biomaterials, health IT, and service
providers (including contract research organizations). The Research Triangle region’s
development, in terms of the number of start-ups established per year, largely mirrors the general
national trend, taken from Bioscan in 2013. While the region lags a few years behind the national
trend, there was significant divergence beginning in the mid 1990s, when the region begins to take-
off. After 2010, while Bioscan indicates a decrease in number of firms founded, the Research
Triangle region’s number continues to increase.1
DATA & DESCRIPTIVES
For this analysis, we use data on the universe of 933 firms that were founded between 1980
and 2016 in the Research Triangle Region. The year 1980 is when the Cohen-Boyer patent (which
allowed patenting of recombinant DNA) was issued, the Bayh-Dole Act was passed, and Genentech
was founded as the first dedicated biotechnology firm. This is an appropriate time to begin an
analysis of industry and ecosystem emergence.2 The data used in the analysis is obtained from the
PLACE: Research Triangle Database, which contains information on the universe of life science
firms founded in the Research Triangle region.3 PLACE: Research Triangle includes information on
firms’ founding, sector, activities, milestones, and founders (Feldman & Lowe 2015). Firm lists were
collected from a variety of sources, including universities and support organization archives. Lists
were compiled, de-duplicated, and each firm individually researched for information on founding,
and data were triangulated to ensure accuracy. Firm founding was verified with filings from the
North Carolina Secretary of State. Our data gathering process has the benefit of including start-ups
which fail early, before most available data sources can capture them, thus mitigating issues of
selection bias. Firms were vetted by local veterans of the life sciences industry, providing us
1 Comparison is available from the authors by request. 2 There were 10 entrepreneurial firms operating in this region prior to 1980 but these firms were arguably not
dedicated life science firms. 3 Data was obtained from the PLACE: RTP database in February of 2019.
6
confidence that our dataset contains the full universe of entrepreneurial life science firms. The firm
is our unit of observation: we match on firm name and prior alias used by the firm.
State-level funding data was obtained directly from the two funding sources: the North
Carolina Biotechnology Center (NC Biotech) and NC IDEA. Both of these sources provide early stage
seed funding. NC Biotech is a quasi-public entity dedicated to promoting biotechnology and the life
sciences in North Carolina with a, “mission to accelerate life science technology-based economic
development through innovation, commercialization, education and business growth,” (NC Biotech
2019). Established in 1981, NC Biotech offers business and employment services, and hosts
meetings and events to connect researchers and other industry stakeholders. Starting in 1989, NC
Biotech Center has continuously provided small loans to new ventures. NC IDEA was created in
2006 to support high growth entrepreneurial firms, and was formed out of the break-up of the
Microelectronics Center of North Carolina. NC IDEA began providing seed grants to entrepreneurial
firms in North Carolina that year, and has continuously made such grants available. In 2015 the
organization converted to a private foundation, and in 2016 made its 100th seed grant. For each
seed grant, we collected the date and award amount.
To measure federal funding, we use SBIR and STTR awards from the Small Business
Administration, combined with awards from NIH. Funding from these sources is focused on
research the firm is performing, and awarded based on evaluations of that research. Dollar amounts
of each award are typically larger than those of the State funding sources, but still much less than
the quantities of private funding a firm might receive. Data were obtained from the NIH Federal
RePORTER and SBIR.gov.
Data on private funding was gathered from CB Insights. Most private funding recorded in CB
Insights comes from VC firms and angel funds. Other sources, added in the past decade, include
start-up competition awards and funding from accelerator participation. This data includes
information on the amount, date, and round of the investment, as well as the name of the investing
entity. For some observations the dollar amount of the investment was missing. If the amount could
not be determined from online press releases or other sources, we used $100,000 as a modest
imputation.
Together state, federal, and private funding accounts for the majority of external finance a
firm will receive.4 Other possible external sources include foundation funding, but these awards are
uncommon. Start-up founders might also “bootstrap” by not taking any type of finance that involves
4 All dollar amounts are inflated to 2017 real dollars.
7
an exchange of equity. Bank loans and credit card debt are other means of financing but these data
are not available.
-------------------------
Insert Figure 2 about here
-------------------------
-------------------------
Insert Figure 3 about here
-------------------------
Of the 933 firms, 179 received national funding, 162 received state funding, and 163
received private funding. A total of 264 firms have received public funding from state or federal
sources. Figures 2 and 3 present annual counts of the number of firms receiving funding and the
dollar amounts received by source, respectively. When examining the number of firms receiving
funding from each source, it is apparent that most firms who have received funding receive it from
the federal government. There is a general upward trend among all funding sources until the
financial crisis, when all three sources allocate funds to less firms, with state funding experiencing
the largest drop. When examining which source provides the most funding in terms of dollar
amount, it is evident that private funding provides the most, followed by national support, and then
state. Thus, even though the federal government supports the largest number of firms, private
finance provides the greatest dollar amount of funding.
-------------------------
Insert Table 1 about here
-------------------------
Finally, we examined sectoral differences in the development of the region’s life science
cluster. Each firm was classified by the authors into one of the following six sectors, based on their
technology focus: (1) dedicated biotechnology or biomaterials, (2) human therapeutics, (3) tools or
diagnostics, (4) medical devices, (5) health IT/informatics, and (6) services. Table 1 provides
definitions of these sectors. Classifications were verified in consultation with local industry experts.
The largest concentration is human therapeutics (357 firms), followed by medical devices (152
firms). Our descriptive analysis, available upon request, revealed that services firms are quite
different than other members of the life sciences industry, receiving their funding largely from
federal sources, with little private sector investment. Furthermore, while these firms are part of the
ecosystem, their activities are more service-oriented rather than research-related. For these
8
reasons, we do not include these firms in the analyses. This reduces our sample of firms from 933 to
821.
METHODOLOGY
Our data allows deep, longitudinal exploration of emergence in one region. This data,
combined with the exploratory nature of our research questions, lends itself to a methodology that
follows an exploratory, empirical analysis. We therefore employ five different techniques to
examine how an industry emerges and develops in a particular local ecosystem, and investigate the
role of start-up finance. Here we present an overview of these methods, which we explicate
mathematically in subsequent sections.
First, appealing to theory on industry and cluster life cycles we examine the emergence and
development of the industry by empirically distinguishing stages. There is no agreed upon way in
the literature to measure different life cycle stages in this context (Wang et al. 2014) so we
introduce the use of threshold regression analysis to help with this task. An exception is the
discriminant method, which Agarwal & Bayus (2002) implement to detect the shift from growth to
maturity.
After we establish the dates upon which the regional ecosystem moved between phases, we
provide a chronicle of the region’s development, relying on detailed qualitative information we
gathered over several years of interviewing entrepreneurs, support organization managers,
entrepreneurial service providers, public officials, and other individuals with first-hand knowledge
of the regional ecosystem. This data provides a qualitative argument for why phase changes
occurred when they did, providing support for quantitative analysis.
Next, we examine how multi-level public and private start-up funding influenced the
region’s development. First, we establish the relationships between federal, state, and private
finance, and examine how these differed among stages using a dynamic random effects probit
model with unobserved heterogeneity and regime-switching analysis. With these two techniques
we are also able to examine how receipt of one type of funding is related to prior funding from the
same source, and other sources, and how this relationship changed across the ecosystem lifecycle.
Third, we examine how funding influences ecosystem development through firm survival.
To answer this question, we employ discrete event history analysis to see how the source, or
combination of funding, influences firm survival. Start-up survival is of paramount importance to
ecosystem development.
9
DETECTING EMERGENCE
The ecosystem life-cycle model builds on the industry and product life cycle theory’s logistic
S-shaped curve characterized by different underlying functions that specify slope or rate of change
(Agarwal & Bayus 2002; Argyres & Bigelow 2007; Moeen & Agarwal 2017). The curve is
characterized by changing functions, with varying degrees of convexity during emergence. When a
system enters a period of decline, the function then becomes concave. Our intent is not to map the
Research Triangle’s ecosystem to the life cycle curve, rather to use the cycle and its derivative
elements to characterize the nonlinearity of ecosystem emergence.
Our method uses the distribution of firms to detect discontinuities in the growth of firms in
the region. Building on derivative elements of the life cycle, we detect the dates when the sign of the
second partial derivative changes with respect to time. We thus write the industry dynamic as a
second order polynomial function of time, that is:
𝑦𝑡 = 𝛼𝑡 + 𝛽𝑡2 + 𝜀𝑡 (1)
which translates into the following first difference model:
∆𝑦𝑡 = 𝑦𝑡+1 − 𝑦𝑡 = 𝛼 + 𝛽(2𝑡 + 1) + 𝑢𝑡 . (2)
We can further simplify this expression by demeaning the first difference model, and obtain:
∆𝑦𝑡 − ∆𝑦𝑡 = 𝛽[(2𝑡 + 1) − (2𝑡 + 1) ] + 𝑣𝑡 (3)
With this expression, we need only estimate one parameter, β, which corresponds to the second
partial derivative of the industry dynamic with respect to time. Consequently, by identifying the
structural change in the β value, we are able to precisely detect different industry life-cycle
transition points and phases.5
We employ a threshold model to detect break points in the regional industry life-cycle.
Hansen (2000) argues that threshold models can capture abrupt breaks or asymmetries observed
in most time series data. Threshold models are flexible in that they allow specification of a known
number of thresholds or let the software find the optimal number of thresholds through different
efficiency criteria (Hansen 2001 provides a survey of threshold regression models).
Consider first our empirical specification for the dynamic of the industry,
∆𝑦𝑡 − ∆𝑦𝑡 = 𝛽𝑥𝑡 + 𝑣𝑡 (4)
5 Additional details are available upon request.
10
where 𝑦𝑡 represents the number of active firms in region at time t. Let 𝑥𝑡 represent our first
difference demeaned variable of time [(2𝑡 + 1) − (2𝑡 + 1) ]. In our case, the empirical specification
does not include any stage-invariant covariates. The dynamics of the industry life-cycle are only a
function of time, which is also our threshold variable. Therefore, we estimate the following
threshold model:
∆𝑦𝑡 − ∆𝑦𝑡 = ∑ 𝑥𝑡𝛽𝑗𝐼𝑗(𝜆𝑗 , 𝑡)𝑛+1
𝑗=1 + 𝜀𝑡 (5)
Here, 𝜆𝑗 represents the threshold value in t, that is, the break point dates in the industry life-cycle.
Estimation uses the threshold command in StataSE 15.1.
-------------------------
Insert Table 2 about here
-------------------------
Threshold Estimation Results
Table 2 presents three different threshold estimations using different specifications. The
typical life cycle specification is 4 stages but we are not sure a priori that the region has
experienced decline. Model 1 set the number of stages to 3 stages while Model 2 specifies 4 stages.
Model 3 uses the Bayesian Information Criteria (BIC) to decide the optimal number of thresholds.
Models 1 and 3 provide identical estimation. According to the BIC, the optimal number of
thresholds is two, which results in three threshold phase changes. This is what was imposed in
Model 1. The results suggest that the Research Triangle’s life science industry has experienced
three different stages from 1980 to 2016. The initial phase, which exhibits a positive convex
dynamic (β=0.445), began in 1980 and lasted until 1997. The next stage is estimated between 1998
and 2009 with a stronger convex dynamic (β=0.899), indicating higher growth. Finally, we estimate
that the industry entered into a third stage in 2010, marked by a significant change in the industry
dynamics as the curve becomes more concave (β=-0.065).
Model 2 imposes four different stages for the industry life-cycle. An important artifact is
that we still detect the same time period for our three main stages: an initial phase (1980-1997),
followed by higher growth (1998-2009), and then slower growth (2010-). By imposing another
threshold, we detect the turning point of the second phase–the date when the dynamic becomes
less convex. Model 2 highlights a strong dynamic during the first part of this stage (1998-2004)
with a β estimated at 1.4767, whereas the second part (2005-2009) has a β estimated at 0.775,
11
which is approximately half the former.6 Overall these results are robust and confirm the existence
of three ecosystem phases from 1980 to 2016.
-------------------------
Insert Figure 4 about here
-------------------------
Qualitative Mapping of Regional Life Cycle Phases
We examine these quantitative results with a qualitative mapping of events in the region.
Over time, the Research Triangle region slowly nurtured an entrepreneurial ecosystem, thanks to
mergers and layoffs from high-profile multinational firms (like GlaxoSmithKline), a more aggressive
technology transfer stance from the region's research universities, and the development of a
plethora of support institutions. The result is a steadily growing number of entrepreneurial firms
and the presence of intermediary entrepreneurial support institutions. Notable events in the
development of the Research Triangle region’s life science industry are chronicled in the timeline in
Figure 4. Specifically, we examine events that preceded the discontinuities that occurred in 1997
and 2009. Within this complex system it is impossible to establish causality, however we are
interested in events that coincide with moving the ecosystem from one phase to the next.
The three universities, University of North Carolina at Chapel Hill (UNC), Duke University
(Duke), and North Carolina State University (NCSU) were important for anchoring the region and
providing access to skilled labor and collaborations. IBM and the National Institute of
Environmental Health Sciences (NIEHS) were early tenants in the Park, moving in 1965. Becton
Dickinson and Wellcome were the first life science firms, becoming Park tenants in the early 1970s.
During the Phase 1, state support of biotechnology strengthened. During the early 1980s
policymakers and civic leaders initiated a concerted effort to build the biotechnology industry in
the region as they believed the region’s research universities could be leveraged for competitive
advantage in knowledge-intensive industries. The universities had already demonstrated
considerable success in bringing in federal research funding but in this time period they established
technology transfer offices to streamline commercialization efforts. Duke and NCSU opened tech
6 Using industry turnover as the dependent variable with sales data is confirmatory of the breaks identified with
industry density presented here, but data problems limit their explanatory power. Therefore, whatever the dependent
variable, it seems that using our empirical specification with threshold regression allow us to precisely determine the
date of the life-cycle transition points. Results are available from the authors upon request. Data on sales was
obtained from the National Establishment Time Series (NETS) for the years 1990-2014 (Walls & Associates 2013).
We supplement NETS with more recent data from Lexis Nexis6 and used Stata’s mipolate command to interpolate
across missing years as this data was usually available for one year between 2014 and 2018 for our firms.
12
transfer offices in the mid 1980s, with UNC following later in the mid 1990s. The Council for
Entrepreneurial Development (CED) and the NC Biotech Center were both founded in 1984. The NC
Biotech Center is a state-funded industry development organization. CED was founded by the
Raleigh Chamber of Commerce, along with several important stakeholders in the industry, to
support high-tech entrepreneurship in the Triangle region. Both organizations are still active today.
During the 1980s, a series of federal policies also helped strengthen researchers’ ability to perform
and commercialize life science research. This deepening sociopolitical legitimation of the new
industry laid important groundwork for its continued development (Aldrich & Fiol 1994)
When Glaxo moved its headquarters to the Triangle in 1983 it was well positioned to take
advantage of possible collaborations with the universities. As large firms moved to the Triangle,
spin-outs from the university started to become more common. Quintiles, now the international
firm IQVIA, was founded by two UNC professors in 1981. Sphinx Pharmaceuticals, a Duke spin-out,
was the region’s first initial public offering (IPO) in this industry in 1991, and the largest in state
history at the time ($70 million). Intersouth Partners, the region’s first VC firm, provided Sphinx’s
seed funding and original incubation space, and Sphinx founders received a $3 million grant that
they used to start the company. Sphinx was ultimately bought by Eli Lilly & Company in 1994. After
this acquisition, Eli Lilly established an RTP presence. Today, Intersouth Partners boasts over 30
companies in the region that can be traced to Sphinx’s founders. This story highlights the deep
connections between incumbents, universities, and spinouts in the Research Triangle region
(Intersouth Partners 2019).
The early success of a few home-grown start-ups (including PPD, Quintiles, Sphinx, Embrex)
brought national attention to the region. Furthermore, the nature of these early start-ups (Quintiles,
PPD) and the attraction of large corporations (Glaxo, Fujifilm Diosynth) meant start-ups had
flexibility to easily outsource activities to high-profile, neighboring contact research organizations
and contract manufacturing organizations, while also exploiting networks within these
organizations and universities. Several IPOs and large VC deals likely increased the amount of
funding going to firms as the Triangle region achieved more notoriety in the life sciences.
These activities laid the foundation for the cluster to expand and move from Phase 1 to
Phase 2. The 1990s saw a takeoff of spinout companies in the Research Triangle region. This trend
of an increasing number of start-ups was enhanced by a series of spin-outs created from former
Glaxo and Burroughs Wellcome employees after layoffs caused by the 1995 merger of the two
companies left employees jobless, but with ideas for new firms. Several more local VC funds were
founded during the 1990s as well, and the decade witnessed more IPOs and some larger VC deals.
13
UNC and NSCU established a joint biomedical engineering department in 2003, just a few years
after NCSU created its Centennial Biomedical Campus with lab and office space for life science start-
ups. Interestingly, the timing of entry into Phase 2 comports with an early prediction made by
Aldrich and Zimmer (1986) that significant entrepreneurial activity would not occur in the
Research Triangle region until the mid 1990s because of the lack of density in social networks and
low start-up rates when considering the time it took for Silicon Valley and Route 128 in Boston to
mature.
The beginning of Phase 3 begins in 2009, which corresponds with the financial crisis. While
we might call this period “Maturity” because of the decrease density of of spin-outs, we actually
cannot be sure. Most life cycle theories attribute the onset of the Maturity stage to a period of
industry “shakeout” due to competition, saturation, or the like (Agarwal et al. 2002). With the onset
of the region’s third phase so closely matching an exogenous economic shock, it is likely that this
break point which brought an apparent downturn only periodically stunted the development of the
region, which may already be experiencing a renewal as new technological advancements are made
that bring new market opportunities. Still, the results indicate a structural break from the prior
period.
The number of start-ups continues to increase annually, and the ecosystem continues to be
successful along other performance measures. During post-recession the region witnessed an influx
of entrepreneurial support organizations including incubators and accelerators from outside the
region. Interestingly these organizations have already experienced much turnover. IPOs are rare
events, but local start-ups consistently go public and the number of IPOs continues to increase in
this stage. A recent acquisition of Bamboo Therapeutics by Pfizer brought much attention. With a
$150 million price mark, the deal had the potential to reach over $600 million depending on how
the start-up technology developed. G1 Therapeutics completed an IPO in 2017 for $105 million,
after receiving $1 million in loans from NC Biotech and over $95 million in venture capital. Other
start-ups’ stories demonstrate the embeddedness of the entrepreneurs and technologies in the
region. Advanced Liquid Logic, which spun-out of Duke, was sold to Illumina in 2013 for $96
million. The founders then licensed the technology back from Illumina and started Baebies, a
company that develops newborn screenings. It is important to note that the number of local VC
firms supporting the life sciences has declined over the past decade, though sources indicate RTP
start-ups are becoming more adept at acquiring non-local VC (personal interview, 2019).
14
-------------------------
Insert Table 3 about here
-------------------------
FUNDING SOURCES & EMERGENCE
While the number of firms funded and the amount provided by federal sources increases
throughout the three stages, the number supported by private sources sees a dramatic increase
from Stage 1 to Stage 2 (see Table 3). State funding is significantly lower than federal and private in
terms of dollar amount but has seen more growth than the other two sources going from Stage 2 to
Stage 3.
Dynamic Random Effects Probit Model with Unobserved Heterogeneity & Regime-Switching
We estimate a dynamic random effects probit model with unobserved heterogeneity to
examine the relationships between funding sources across the entire panel and then implement a
regime-switching analysis on the same model to see how these results change depending on the
stage. We perform this analysis separately for each funding source (state, federal, and private).
Table 3 demonstrates variation across funding sources for the three stages
The dynamic random effects probit model is appropriate for several reasons. First, our
setting is dynamic where lagged values of funding from each source are needed to estimate the
probability of funding from that source in the current time period. We are most interested in the
probability of securing funding, given the prior distribution of sources. As such, we must account
for state dependent processes. Second, we do not use the value of funding because we are more
interested in the links between funding sources in the ecosystem and want to avoid a value-effect
from funding caused by noise related to large or erratic values. Thus we have a binary outcome
variable. For example, if a start-up received funding from the same source in every year, but the
amount decreased we would see a negative effect even if the probability of receiving funding is
high. Third, the relationships between state, federal, and private funding are nonrecursive. It is for
this reason we cannot estimate for each type of funding simultaneously. Due to the nonrecursive
and binary nature of the data, a structural equation model is inappropriate. We follow Grotti &
Cutuli (2018) and Rabe-Hesketh & Skrondal (2013) in estimating a model of the following form:
𝑦𝑖𝑡∗ = 𝜌𝑦𝑖𝑡−1 + 𝛾𝑍𝑖𝑡 + 𝑐𝑖 + 𝑢𝑖𝑡 (6)
15
Here 𝑦𝑖𝑡∗ is the probability of firm 𝑖 receiving funding at time 𝑡. It is a function of funding received in
the previous period, 𝑦𝑖𝑡−1, which accounts for state dependence, and covariates 𝑍𝑖𝑡 , which control
for firm-level bias. 𝑐𝑖 is the firm-specific unobserved effect, while 𝑢𝑖𝑡 is an idiosyncratic error term.
The firm-specific unobserved effect accounts for heterogeneity between firms and is written
as:
𝑐𝑖 = 𝛼0 + 𝛼1𝑦0 + 𝛼2�� + 𝛼3𝑋0 + 𝛼4 (7)
While 𝛼0 and 𝛼4 (the firm-specific time-constant error) are purely random, 𝑦0 represents the initial
value of the dependent variable, while 𝑋0 is a row matrix of the initial value of the time varying
explanatory variables for all firms, 𝑋0 = {𝑋01, 𝑋02, … , 𝑋0𝑛}. Here, this is the initial value of the other
two sources of funding. For example, if 𝑦𝑖𝑡∗ is the probability a firm received private funding in time
𝑡, 𝑋0 accounts for the initial value of state and federal funding. Similarly, �� is a row matrix of the
mean value of the time-varying covariates (in the example, state and federal funding) for all firms,
�� = {��1, ��2, … , ��𝑛}. Specifically, it is the within-firm average based on all time periods. 𝑐𝑖then
accounts that the probability of receiving one type of funding is likely influenced by other funding
sources in prior periods (Grotti & Cutuli 2018; Rabe-Hesketh & Skrondal 2013; Wooldridge 2005).
The 𝑋 are a subset of 𝑍. All models are heteroskedastic robust.
The dynamic random effects probit model assesses state dependence in the probability of
receiving funding but does not account for differences across life cycle stages. We estimate a
regime-switching regression analysis, which is a generalization of the linear model that accounts
for temporal breaks in the effects of the variables (Elhorst & Fréret 2009; Montmartin et al. 2018).
We expect the influence on funding is not linear with time and therefore consider the break points
we identified in the threshold analysis. The model allows us to estimate stage-specific coefficient
values.
We estimate one dynamic random effects probit model with regime-switching for each
funding source, including stage specific indicators and interactions that represent the transitions.
For example, the model for private funding is as follows:
𝑃𝑅𝐼𝑉𝑖𝑡 = 𝛼1 + 𝛽11𝑃𝑅𝐼𝑉𝑖𝑡−1 + ∆𝛽12𝑃𝑅𝐼𝑉𝑖𝑡−1 + ∆𝛽13𝑃𝑅𝐼𝑉𝑖𝑡−1 + 𝛽21𝑆𝑡𝑎𝑡𝑒𝑖𝑡−1 + ∆𝛽22𝑆𝑡𝑎𝑡𝑒𝑖𝑡−1
+ ∆𝛽23𝑆𝑡𝑎𝑡𝑒𝑖𝑡−1 + 𝛽31𝐹𝑒𝑑𝑒𝑟𝑎𝑙𝑖𝑡−1 + ∆𝛽32𝐹𝑒𝑑𝑒𝑟𝑎𝑙𝑖𝑡−1 + ∆𝛽33𝐹𝑒𝑑𝑒𝑟𝑎𝑙𝑖𝑡−1 + ∆𝛼2
+ ∆𝛼3 + 𝑍𝑖𝑡 + 𝑐𝑖 (8)
Here, the 𝛽1𝑖 represent the focal funding source and coefficients of key interest (private funding in
this equation), with 𝛽2𝑖 and 𝛽3𝑖 representing lagged values of the other two funding sources. The
𝛽𝑖1represent coefficients for Stage 1, which we choose to be the base effect, while the ∆𝛽𝑖2 represent
the change of the effect from stages 1 to 2 and the ∆𝛽𝑖3 represent the changing effect in stage 3 with
16
respect to stage 1.7 For the regime analysis, the 𝑐𝑖 unobserved effect covariates include only state
and federal funding. All models are heteroskedastic robust.
-------------------------
Insert Table 4 about here
-------------------------
-------------------------
Insert Table 5 about here
-------------------------
-------------------------
Insert Table 6 about here
-------------------------
Our estimation includes several firm and industry level covariates (𝑍𝑖𝑡). These are listed in
Table 4 with the literature that included these control variables. We include firm-level controls for
founding team size, founder human capital, sector, and establishment date, as well as the region-
level industry density. Covariates describing founders’ human capital are commonly used controls
for start-up quality, since start-ups generally lack data on other metrics such as sales that would
measure this. Summary statistics and correlations are provided in Tables 5 and 6.8
-------------------------
Insert Table 7 about here
-------------------------
Dynamic Random Effects Probit Estimation with Regime Switching Results
Results from the dynamic random effects probit estimation are provided in Table 7, while
Tables 8, 9, and 10 present the dynamic random effects probit result with regime switching
analysis. We present tables showing the total effects and their significance by stage. The full table of
regime-switching estimation results can be found in the Appendix Table A1.
7 We calculate the net effect by stage by adding the coefficient of the base effect and the coefficient of the associated
change for stages 2 and 3. We calculate standard errors using the covariance matrix of the coefficients and calculate
a t-statistic to discern statistical significance at the usual levels. 8 For 31 firms it was impossible to determine values for some of these variables. Eight firms are missing all founder
data. The other 23 firms had founders with some missing details of their work experience and/or educational
background. For these firms we used either the minimum or average value to replace the missing values, whichever
would cause any bias to be in the direction a null funding. In the case of the same prior industry experience variable,
we inflated the count by how many founders there are in the founding team for firms for which only some founders
were missing this variable.
17
Table 7 demonstrates that, even without considering the influence of life cycle stage, there
are differences in the dynamic relationships between funding sources. Model 4 considers the
influence of lagged indicators for whether a firm received private, state, and/or federal funding
(time t-1) on the probability of receiving private funding in time t. We detect a positive influence of
private funding (t-1) and a positive influence of state funding (t-1) on private funding in time t. We
do not, however, find an influence of federal funding in time t-1 on private funding in time t. These
findings indicate overall that private sources may use state funding as a signal. Furthermore, prior
private funding is a predictor of future private funding. There is no influence of any type of funding
in the prior time period (t-1) on state funding in the current period (time t) (Model 5). We argue
this provides confirmation of what we know about state funding—that it is awarded at a very early
stage, usually before other funding sources are available to firms, and that it is rare that state grants
and loans are awarded to the same firm more than one time. Finally, in Model 6, considering the
dynamic random effects probit model across the whole panel for the influence on federal funding in
time t, we detect a positive effect of federal funding (t-1), a positive effect of state funding (t-1) and
a positive effect of private funding (t-1). This means firms federal sources rely on signals from other
funding sources and invest in firms likely that are nearing the end of seed-stage. Furthermore,
federal funds are likely to go to firms that have already received some federal funding.9
-------------------------
Insert Table 8 about here
-------------------------
-------------------------
Insert Table 9 about here
-------------------------
-------------------------
Insert Table 10 about here
-------------------------
The stage based model results in Table 8 provide a deeper understanding of these
relationships for private funding. We find that the positive effect of private funding in time t-1 on
private funding in time t is only present during Stage 1 and disappears in later stages. Interestingly,
the positive influence of state funding is present in Stages 1 and 3. By Stage 3, state funding sources
are encouraging firms to go after private sources and using their own networks to help the firms
9 While the federal SBIR/STTR programs grant funds in phases, with Phase II and Phase III follow-on on funds, this
structuring is not sufficient to explain this finding as it is unlikely a firm will receive a Phase I grant in year t-1 and
Phase II in year t. Firms are receiving different grants for different projects.
18
make connections and access new funding sources. The impact of federal funding is not strong. We
detect no effect in Stage 1 and only a slightly negative effect during Stages 2 and 3. This finding
could be driven by a few firms that rely on federal research grants to fund their work rather than
trying to attract private funding—a practice which has earned the nickname “SBIR mills.”
When considering North Carolina State funding in time t (Table 9), we find no influence of
prior state funding or prior federal funding in year t-1, regardless of life cycle stage. This finding
therefore most closely aligns with the results for the full panel. We do, however, see a positive
influence of private funding in year t-1 in Stage 3. As the private funding mechanism has evolved
over time we have begun to see more angel funding and more sources of private funding being
directed to early stage firms, such as accelerators and competitions. It then makes sense why
private seed funding might pre-date public seed funding and even act as a signal to state-level
funding sources for more recent start-ups. We know from our own experiences and interviews in
the region that state agencies which support biotech devote considerable attention to developing
networks and relationships with local VC and angels.
If we consider federal funding in time t (Table 10), we see it is strongly influenced by all
three sources in time t-1 during the first stage. While this influence drops for private funding in the
second stage, it is maintained for state and federal funding. However, the influence of federal and
state funding in time t-1 on federal funding in time t disappears in the third stage. This indicates
that early in the emergence process federal funding may rely heavily on signals from other funding
sources and be more likely to re-invest in firms which show promise. As the industry matures,
however, federal funding sources may be more likely to be directed to firms developing more high-
risk and untested technologies which private may not be interested in supporting and state does
not have the bandwidth to support..
Overall these results demonstrate it is important to consider the wide variation in the
relationships between funding sources, considering each source separately and considering the life
cycle stage. Interestingly as well, we find across all regime models that the initial value of other
funding does not well explain future funding, while the mean value is consistently statistically
significant.
FUNDING SOURCES & FIRM SURVIVAL ACROSS THE ECOSYSTEM LIFE CYCLE
We develop a complementary loglog (cloglog) model including non-parametric frailty and a
logarithmic duration dependence to model the hazard rate for survival of the Research Triangle
19
region entrepreneurial life science firms. Our survival analysis begins in 1980 and ends in 2016
with an interval of one year for grouping. Therefore, we cannot consider that the ratio of the
interval length to the typical spell length is low, as many entrepreneurial life science firms die in
their first 5-10 years. Consequently, though survival occurs in continuous time our data are
grouped and the observations on the transition process are summarized discretely rather than
continuously. Thus we use econometric methods dealing with discrete survival time data and a
complex dynamic for the hazard rate concerning entrepreneurial life science firms.
The cloglog model is best suited for our data (see Jenkins 2005). The discrete time hazard
rate is modeled as:
ℎ(𝑡, 𝑋) = 1 − 𝑒𝑥𝑝[− exp(𝛽′𝑋 + 𝛾𝑡)] (9)
where X are the covariates influencing the hazard rate and 𝛾𝑡 summarize the pattern of duration
dependence in the interval hazard. There are two main ways to model 𝛾𝑡, either by not placing any
restrictions on how the 𝛾𝑡 vary from interval to interval (in which case the model is a type of semi-
parametric one) or by specifying a parametric functional form. The most common duration
dependence specifications are the logarithm dependence [𝛾𝑡 = 𝑙𝑜𝑔(𝑡)], polynomial dependence
[𝛾𝑡 = 𝑧1𝑡 + 𝑧2𝑡2 + ⋯ + 𝑧𝑝𝑡𝑝], or assume that 𝛾𝑡 is piecewise constant, that is groups of years are
assumed to have the same hazard rate but the hazard differs between these groups. In order to
investigate this point, we take the hazard rate value obtained from the life-table method and test
which of these three duration dependence specifications is able to explain the hazard rate.
Efficiency criteria indicate the polynomial dependence is most useful to model the duration
dependence, followed by logarithm dependence and the piecewise constant dependence.
We include the same covariates used in the prior analysis. Another important consideration
is the potential presence of unobserved heterogeneity in the hazard rate. We may face some
problems with omitted variables. Indeed, we do not have, for instance, employment data that would
consider the size of the entrepreneurial firm. And as shown by Bruderl and Schussler (1990), Ortiz-
Villajos and Sotoca (2018), and others, size matters for the survival of entrepreneurial firms. Thus,
we need to take into account the presence of unobserved heterogeneity, or “frailty”, as it is called in
the literature.
We thus now consider the following generalized cloglog model:
ℎ(𝑡, 𝑋|𝑢) = 1 − 𝑒𝑥𝑝[− exp(𝛽′𝑋 + 𝛾𝑡 + 𝑢)] (10)
where 𝑢 represents the unobserved individual effects. There are three main ways to model these
unobservable variables: using a normal distribution [𝑢~𝒩(0, 𝜎2)], a gamma distribution
[𝑢~Γ(1, 𝜎2)] or a non-parametric approach which fits an arbitrary distribution using a set of
20
parameters. These parameters comprise a set of mass points and the probabilities of a person being
located at each mass point. The graph showing the hazard rate is not helpful to discriminate
between these three possibilities.10 Consequently, we choose the best specification for frailty by
comparing the relative performance of each estimation.
-------------------------
Insert Table 11 about here
-------------------------
Discrete Event History Analysis Results
Our first survival estimation results can be found in Table 11. Information criteria from
estimations comparing the normal distribution, non-parametric approach, and gamma distribution
indicate the non-parametric approach is preferred. We present results from the non-parametric
estimation only. For comparison, results from the Gaussian and gamma estimations can be found in
Appendix Tables A2 and A3, respectively. The robustness of this model is confirmed by the similar
results found when modeling heterogeneity with Gaussian and gamma distributions. Table 11
presents four models, each with a different functional form of the main dependent variable. Model
16 displays the results using the logged cumulative dollar amount of funding a firm receives
separately from state, federal, and private sources. Model 17 estimates instead the logged
cumulative number of investments a firm receives from each source separately. Model 18 again
uses the logged cumulative dollar amount invested and adds a squared term for each funding
source. Model 19 presents the preferred specification, which uses the logged cumulative dollar
amount, with a nonlinear effect of private funding.
Overall, we find that firms that received federal funding experience less chance of failure,
and this effect is nonlinear with the cumulative amount received. Importantly, the effect of federal
funding is statistically significant across all specifications. The value of the coefficient for federal
funding in Model 19 indicates that if federal funding is increased by 1%, the probability of failure is
reduced by 0.034%, which provides support for the argument that federal funding plays a role in
industry emergence. This funding helps firms extend research lines by hiring staff or outsourcing.
Of course these grants depend on the quality and promise of the research, but by modeling the
unobserved heterogeneity in our discrete time models we alleviate concerns of cross-firm
heterogeneity.
10 Additional details are available upon request.
21
We also find that firms that received private funding experience less chance of failure, and
the effect is non-linear with the cumulative amount received. In increase of 1% in private funding
reduces the risk of failure by 0.044% which is slightly larger than the influence of federal funding.
This result confirms prior literature on the importance of VC and other private funding sources to
emerging ecosystems. Private funding generally comes in larger amounts than state and federal,
though as mentioned previously such trends have changed, and focuses on building the firm.
Private therefore should help firms survive and have a larger magnitude influence than federal
funding. Private funding from VC and angels generally comes with mentorship and accesss to the
funders networks of resources which further help firms develop.
Finally, however, we find that firms that received North Carolina state funding do not
experience less of a chance of failure, as results are not significant, controlling for the amount of
federal and private funding. This finding does not suggests state funding is inconsequential. Instead,
state funding is likely to impact firms in many ways, not just on their ultimate survival. The smaller
dollar amount of state funding may be too small to have an ultimate impact on survival. As show in
previous results, state funding does influence the probability of later private and federal funding,
which do show a positive relationship to firm survival.
Concerning controls, a higher level of competition in the industry and region, as measured
by industry density, has a significant effect on the risk of failure, decreasing the risk of failure. In
contrast, being a newer firm with a later establishment actually increases the probability of failure.
The fact that a firm has an entrepreneur that has already worked in the same industry decreases
the risk of failure. Also, firms with more co-founders have a decreased probability of failure. Finally,
firms who have more serial entrepreneurs have an increased probability of failure. This finding is at
odds with theory about the benefits of prior entrepreneurial experience but could suggest that
these entrepreneurs are better at knowing how to fail fast.
-------------------------
Insert Table 12 about here
-------------------------
We examined further variations on the functional forms of the state, federal, and private
funding variables. These results are presented in Table 12 (Gaussian and gamma specifications in
Appendix Table A4). Here we include all possible varieties of funding that a firm might receive as a
series of indicator variables, with the reference category being firms that received no support.
These results confirm the previous ones and improve our understanding of the effect of funding an
emerging ecosystem. Receiving only private funding or any combination of sources decreases the
22
probability of failure. Receiving only state funding or only federal funding has no effect, while
receiving state and federal funding, even without private funding, does decrease the probability of
failure. However, an important finding from these models is that firms that received both private
and federal funding and firms that received all three sources of funding have the lowest magnitude
chances of failure.
When examining the controls, we again see that being a newer firm in the ecosystem
increases the probability of failure. Results also indicate that firms whose founder(s) have some
prior industry work experience have a lower chance of failure, as well as firms with team founded
firms. Again we see that firms who have more serial entrepreneurs have an increased probability of
failure.
-------------------------
Insert Table 13 about here
-------------------------
Discrete Event History Analysis with Regime-Switching
We also implement a regime-switching regression analysis to investigate the stage-specific
influence of each funding source on the survival prospects of firms. The results are provided in
Table 13. We apply the same regime method here as for the prior dynamic probit model and
calculate stage specific coefficients the same way. The full estimation can be found in Appendix
Table A5.
Findings indicate there exist important differences in the influence of funding on firms’
probability of failure across ecosystem life cycle phases. This is an important finding as it indicates
looking at the industry over the entire time period overlooks important differences across regimes.
In terms of private funding the regime findings accord with our previous finding that private
funding is not significant overall in decreasing the probability of firm failure without a square term
(see Table 11 for reference), because this effect is changing over time due to industry dynamics. In
Stage 1, private funding is not helping firms survive. This could be due to the fact that these firms
were entering a new industry and may have been very high risk. Furthermore, the private funders
may not have yet developed expertise in identifying more promising technologies and start-ups as
dominant forms in the new industry are still undiscovered. In Stage 2, private funding becomes
associated with a decreased probability of firm failure, but this is not statistically significant. In the
third stage (Model 23), however, we see a decreased probability of failure and this is stronger in
magnitude and statistically significant. In contrast, federal funding is associated with a decreased
23
probability of firm failure across all stages. The statistical strength of this effect decreases in the
third stage, but the magnitude increases. This means federal funding decreases the probability of
firm failure in Stage 3 with a higher probability than in Stages 1 and 2. When considering state
funding we do not see an influence until Stage 3, when the funding begins to reduce the probability
of firm failure.
These results demonstrate that all three funding sources become more efficient in their
funding allocations over time so that by the third phase these sources are all associated with a
decrease in firm failure above prior stages. Ultimately, this method improves our understanding of
how the funding that is available to firms in an ecosystem is related to their survival, and how this
influence differs depending on industry dynamics which define the ecosystem life cycle.
DISCUSSION & CONCLUSION
Surprisingly little research examines how financing that supports entrepreneurial
businesses indeed promotes regional growth and influences the emergence of new industries.
Studies tend to examine one program in isolation, either venture capital or specific government
programs, with a focus on the degree to which the receipt of that particular funding affects firm
performance. Yet the combined impact on a regional economy—the sum of effects on individual
firms—remains unexplored. There is general consensus that the social rate of return to public R&D
investment is high (Toole 2012), which is used as a justification for government funding. Another
related stream of literature considers the degree to which public funding affects private funding or
the extent of crowding-out versus crowding-in. This topic has been debated since the 1960s (see
Zuniga-Vicente et al. 2014, for a survey of this vast empirical literature). The empirical evidence
suggests that government R&D subsidies lead to an additionality effect and induce private
investment at the firm level. For example, Howell (2017) finds that a government Small Business
Innovation Research (SBIR) award doubles the likelihood of the subsidized firm to subsequently
receive venture capital. The literature is relatively silent on the relationship of government
financing to other sources of financing at the aggregate regional level.
We find emergence and the ecosystems view of the region to be a useful extension of the
theory for explaining how regions, through myriad small efforts of firms and ecosystem
stakeholders, develop capabilities through public and private financial resources that help them
grow and develop new industries. In order to empirically explicate the region’s development, we
24
employ several new techniques which allow us to assess varying temporal relationships within the
region over a long-time horizon.
First, through threshold analysis we find that the Research Triangle region’s life science
cluster has experienced two distinct ecosystem life cycle transition points corresponding to three
phases. The first phase, lasting until 1997 coincides with a period of capability building dominated
by state and federal sources. Then, the region enters a period of dramatic growth characterized by a
higher number of start-ups entry and the availability of greater amounts of private finance as the
region gained prominence in the US. This period of growth began to slow with the start of the
financial crisis as shown by the industry dynamic beginning to exhibit a concave dynamic with
respect to time. Time will tell whether this is actually a downturn for the region or rather the result
of an exogenous economic shock if the start-up entry rates regain momentum. Our contribution is
not mapping the life cycle stages onto an ecosystem, rather we are arguing that an ecosystem goes
through different phases and funding has a different influence over the phases.
Second, the dynamic random effects probit model depicts the relationships between
funding sources at the firm level over time. We find prior state and private funding (time t-1)
influences private funding in time t, while no prior sources influence state funding in time t.
However, we see a relatively strong influence of prior funding from all three sources on federal
funding in time t. Combined with regime analysis, we are able to show that these relationships
actually differ across different life cycle stages of the ecosystem. In the initial phase of emergence,
private funding is strongly influenced by prior private and state funding, while federal funding is
strongly influenced by prior funding from all three sources. Moving to a phase of higher growth we
see a consistent influence of prior federal and state funding on federal spending, but we lose the
influence of private and state funding on state spending. The trends change again during the third
phase, when state funding actually influences private spending, a finding that would be overlooked
with an analysis of the full panel alone. We also find the converse, that prior private funding
influences state spending.
Finally, discrete event history analysis provides evidence that the mix of multi-level public
and private funding of these firm matters for firm survival, and therefore ecosystem emergence.
Necessarily, firms in an emerging industry much be able to reach a point of viability in order for an
ecosystem to develop around them. Notably our results show the importance of having a strong
base of federal funding. Private funding, which provides substantially more funds to Research
Triangle firms in terms of dollar amount than both public sources combined, also decreases the
chances of firm failure. Contrary to expectations, we do not find an effect for state funding, which
25
typically provides the least funding in terms of dollar amount. Still, our results show the
interrelationship between funding sources as firms that receive funding from a variety of sources
experience the greatest magnitude decrease in probability of firm failure. Furthermore, findings
from the application of the regime model to the survival analysis have direct policy implications
and improve the overall survival model. We find an increased efficiency of state funding and private
funding over time, and a consistent influence of federal funding.
Our results shed light on the process by which new industries and ecosystems develop and
emerge in a regional context. They have direct implications for policymakers and for firms. The
results suggest the firms may benefit from strategically considering the sources of funding they
seek depending on the phase of the ecosystem. Furthermore, policymakers interested in growing
their economies by the development of a new industry ecosystem should not merely copy the
current actions of already successful regions, as they would essentially be following the old idiom of
comparing apples and oranges. Our results show that the industry dynamics of the Research
Triangle ecosystem changed in important ways over the ecosystem life cycle. The private, state, and
federal funding sources exhibited differing relationships to each other depending on the stage, and
firm survival was more influenced by different sources also depending on the stage. Consequently,
the paper also makes a methodological contribution to the literature by using innovative methods
to show how results from a full panel mask these dynamics.
26
REFERENCES
Adams, S., (2017). Before the Garage: The Innovation System that Produced Silicon Valley, working
paper, November 30.
Agarwal, R., and Bayus, B. L., (2002). The market evolution and sales takeoff of product
innovations. Management Science, 48(8): 1024-1041.
Agarwal, R., Sarkar, M. B., and Echambadi, R., (2002). The conditioning effect of time on firm
survival: An industry life cycle approach. Academy of Management Journal, 45(5): 971-994.
Agarwal, R., Moeen, M. and Shah, S. K., (2017). Athena's birth: Triggers, actors, and actions
preceding industry inception. Strategic Entrepreneurship Journal, 11: 287-305.
Aksaray, G., and Thompson, P. (2017). Density dependence of entrepreneurial dynamics:
Competition, opportunity cost, or minimum efficient scale? Management Science, 64(5):
2263-74.
Aldrich, H. E., and Fiol, C. M. (1994). Fools rush in? The institutional context of industry creation.
Academy of Management Review, 19(4): 645-670.
Aldrich, H. E., and Zimmer, C. (1986). Entrepreneurship through Social Networks. In D. Sexton, and
R. Smilor (Eds.), The Art and Science of Entrepreneurship: pp 3-23. New York: Ballinger.
Argyres, N., and Bigelow, L., (2007). Does transaction misalignment matter for firm survival at all
stages of the industry life cycle? Management Science, 53(8): 1332-1344.
Bathelt, H., and Boggs, J. S., (2003). Toward a reconceptualization of regional development paths: Is
Leipzig's media cluster a continuation of or a rupture with the past? Economic Geography,
79(3): 265-93.
Blume-Kohout, M.E., Kumar, K.B. and Sood, N., (2015). University R&D funding strategies in a
changing federal funding environment. Science and Public Policy, 42(3): 355-368.
Boden Jr, R.J., and Nucci, A.R., (2000). On the survival prospects of men's and women's new
business ventures. Journal of Business Venturing, 15(4): 347-362.
Bruderl, J., and Schussler, R., (1990). Organizational mortality: The liabilities of newness and
adolescence. Administrative Science Quarterly, 35(3): 530-547.
CBRE Research, (2019). 2019 U.S. Life Sciences Clusters. https://www.cbre.us/research-and-
reports/US-Life-Sciences-Clusters-2019, retrieved July 22, 2019.
Charbit, C., (2011). Governance of public policies in decentralised contexts: The multi-level
approach. OECD Regional Development Working Papers, 2011/04, OECD Publishing.
Chen, K., and Marchioni, M., (2008). Spatial clustering of venture capital-financed
27
biotechnology firms in the US. Industrial Geographer, 5(2).
David, P.A., and Hall, B.H., (2000). Heart of darkness: Modeling public–private funding
interactions inside the R&D black box. Research Policy, 29(9): 1165-1183.
David, P.A., Hall, B.H., and Toole, A.A., (2000). Is public R&D a complement or substitute for
private R&D? A review of the econometric evidence. Research Policy, 29(4-5): 497-529.
Delmar, F., and Shane, S., (2004). Legitimating first: Organizing activities and the survival of
new ventures. Journal of Business Venturing, 19(3): 385-410.
Dencker, J.C., Gruber, M., and Shah, S.K., (2009). Pre-entry knowledge, learning, and the
survival of new firms. Organization Science, 20(3): 516-537.
Drover, W., Busenitz, L., Matusik, S., Townsend, D., Anglin, A., and Dushnitsky, G., (2017). A
review and road map of entrepreneurial equity financing research: Venture capital,
corporate venture capital, angel investment, crowdfunding, and accelerators. Journal of
Management, 43(6): 1820-1853.
Durufle, G., Hellmann, T., and Wilson, K., (2018, May). Catalysing entrepreneurship in and
around universities. Said Business School Research Papers, 2018-13.
Elhorst, J. P., and Fréret, S. (2009). Evidence of political yardstick competition in France using a
two-regime spatial Durbin model with fixed effects. Journal of Regional Science, 49: 931–
951.
Essletzbichler, J., and Rigby, D. L., (2007). Exploring evolutionary economic geographies. Journal of
Economic Geography, 7(5): 549-571.
Feldman, M., Francis, J., and Bercovitz, J., (2005). Creating a cluster while building a firm:
Entrepreneurs and the formation of industrial clusters. Regional Studies, 39(1): 129-141.
Feldman, M., and Lowe, N., (2011). Restructuring for resilience. Innovations, 6: 129-146.
Feldman, M., and Lowe, N., (2015). Triangulating regional economies: Realizing the promise of
digital data. Research Policy, 44(9): 1785-1793.
Foster, J., and Metcalfe, J.S., (2012). Economic emergence: An evolutionary economic
perspective. Journal of Economic Behavior & Organization, 82: 420-432.
Goldfarb, B., Kirsch, D., and Shen, A. (2012). Finance of new industries. In D. Cumming (Ed.), The
Oxford Handbook of Entrepreneurial Finance, pp. 9–44. Oxford, U.K.: Oxford University Press.
Goldstein, J., (1999). Emergence as a construct: History and issues. Emergence, 1(1): 49-72.
Grotti, R., and Cutuli, G., (2018). xtpdyn: A community-contributed command for fitting dynamic
random-effects probit models with unobserved heterogeneity. The Stata Journal, 18(4):
844–862.
28
Hansen, B. E., (2000). Sample splitting and threshold estimation. Econometrica, 68(3): 575-603.
Hansen, B. E., (2011). Threshold autoregression in economics. Statistics and its Interface, 4(2): 123-
127.
Hardin, J. W., (2008). North Carolina’s Research Triangle Park. In W. Hulsink and H. Dons (eds.)
Pathways to High-tech Valleys and Research Triangles: Innovative Entrepreneurship,
Knowledge Transfer and Cluster Formation in Europe and the United States, pp. 27-51.
Netherlands: Springer.
Harper, D.A., and Endres, A.M., (2012). The anatomy of emergence, with a focus on capital
formation. Journal of Economic Behavior & Organization, 82: 352-367.
He, J., and Fallah, M. H., (2009). Is inventor network structure a predictor of cluster evolution?
Technological Forecasting & Change, 76: 91-106.
Howell, S.T., (2017). Financing innovation: Evidence from R&D grants. American Economic
Review, 107(4): 1136-64.
Hyytinen, A., Pajarinen, M. and Rouvinen, P., (2015). Does innovativeness reduce startup
survival rates? Journal of Business Venturing, 30(4): 564-581.
Iammarino, S., and McCann, P., (2006). The structure and evolution of industrial clusters:
Transactions, technology and knowledge spillovers. Research Policy, 35: 1018-36.
Intersouth Partners, (2019). Success Story: Sphinx Pharmaceuticals. Retrieved from
https://www.intersouth.com/sphinx-pharmaceuticals [Obtained Mar. 1, 2019]
Jenkins, S.P., (2005). Survival Analysis. Unpublished manuscript, Institute for Social and
Economic Research, University of Essex, Colchester, UK.
Kalleberg, A.L., and Leicht, K.T., (1991). Gender and organizational performance:
Determinants of small business survival and success. Academy of Management
Journal, 34(1): 136-161.
Karniouchina, E. V., Carson, S. J., Short, J. C., and Ketchen Jr, D. J., (2013). Extending the firm vs.
industry debate: Does industry life cycle stage matter? Strategic Management Journal, 34(8):
1010-18.
Kenney, M., and von Burg, U., (2000). Institutions and Economies: Creating Silicon Valley. In
Understanding Silicon Valley: The Anatomy of an Entrepreneurial Region, edited by M.
Kenney, pp. 218–240. Stanford, CA: Stanford University Press.
Kolympirisa, C., Kalaitzandonakes, N., and Miller, D., (2014). Public funds and local
biotechnology firm creation. Research Policy, 4(3): 121-137.
Krugman, P., (1996). The Self-Organising Economy. Blackwell: Oxford.
29
Lanahan, L., Graddy-Reed, A., and Feldman, M.P., (2015). The domino effects of federal
research funding. PLoS ONE, 11(6): 1-23.
Lécuyer, C., (2006). Making Silicon Valley: Innovation and the Growth of High Tech, 1930-1970.
MIT Press.
Lee, Y.S., (2018). Government guaranteed small business loans and regional growth. Journal
of Business Venturing, 33: 70-83.
Lee, R.D., and Carter, L.R., (1992). Modeling and forecasting US mortality. Journal of the
American Statistical Association, 87(419): .659-671.
Lerner, J., (1996). The government as venture capitalist: The long-run effects of the SBIR
program (No. w5753). National Bureau of Economic Research.
Lewes, G.H., (1875). Problems of Life and Mind, Vol. 2. Kegan Paul, Trench, Turbner: London.
Leyden, D. P., and Link, A. N., (2011, Nov.). Collective Entrepreneurship: The Strategic
Management of Research Triangle Park. Working Paper 11-18.
Luger, M.I., and Goldstein, H., (1991). Technology in the Garden: Research Parks and Regional
Economic Development. Chapel Hill, NC: University of North Carolina Press.
Malakooti, B., (2013). Operations and Production Systems with Multiple Objectives. John Wiley &
Sons: Hoboken, NJ.
Markusen, A., (1996). Sticky places in slippery space: A typology of industrial districts. Economic
Geography, 72: 293–313.
Marshall, A., (1920). Principles of Economics (8th edition). Philadelphia: Porcupine Press.
Martin, R., and Sunley, P., (2011). Conceptualizing cluster evolution: Beyond the life cycle model?
Regional Studies, 45(10): 1299-1318.
Martin, R., and Sunley, P., (2012). Forms of emergence and the evolution of economic landscapes.
Journal of Economic Behavior & Organization, 82: 338-351.
Maskell, P., and Malmberg, A., (2007). Myopia, knowledge development and cluster evolution.
Journal of Economic Geography, 7(5): 603-18.
McCorkle, M., (2012). History and the “New Economy” Narrative: The Case of Research Triangle
Park and North Carolina's Economic Development. Journal of the Historical Society, 12(4),
479-525.
Menzel, M.P., and Fornhal, D., (2010). Cluster life cycles—dimensions and rationales of cluster
evolution. Industrial and Corporate Change, 19(1): 205-38.
Minniti, A., and Venturini, F., (2017). The long-run growth effects of R&D policy. Research
Policy, 46: 316-326.
30
Moeen, M. and Agarwal, R., (2017). Incubation of an industry: Heterogeneous knowledge bases and
modes of value capture. Strategic Management Journal, 38(3): 566-587.
Montmartin, B., and Massard, N., (2015). Is financial support for private R&D always
justified? A discussion based on the literature on growth. Journal of Economic Surveys,
29(3): 479-505.
Montmartin, B., Herrera, M., and Massard, N., (2018). The impact of the French policy mix on
business R&D: How geography matters. Research Policy, 47(10): 2010-2027.
Muscio, A., Quaglione, D. and Vallanti, G., (2013). Does government funding complement or
substitute private research funding to universities? Research Policy, 42(1): 63-75.
NC Biotech. (2019). Home / About Us. [Online]. Retrieved from https://www.ncbiotech.org/about
[Obtained Jul. 23, 2019].
Ortiz-Villajos, J.M., and Sotoca, S., (2018). Innovation and business survival: A long-term
approach. Research Policy, 47(8): 1418-1436.
Potter, A., and Watts, H.D. (2011). Evolutionary agglomeration theory: Increasing returns,
diminishing returns, and the industry life cycle. Journal of Economic Geography, 11: 417-55.
Rabe-Hesketh, S., and Skrondal, A., (2013). Avoiding biased versions of Wooldridge’s simple
solution to the initial conditions problem. Economics Letters, 120(2): 346-349.
Samila, S., and Sorenson, O. (2010). Venture capital as a catalyst to commercialization. Research
Policy, 39: 1348–1360.
Schumpeter, J. A., (1911/34). Theory of Economic Development [1911]. Harvard University
Press: Cambridge, MA.
Schumpeter, J. A., (1942). Capitalism, Socialism, and Democracy. Harper & Row: New York.
Smilor, R., O'Donnell, N., Stein, G., & Welborn III, R. S. (2007). The research university and the
development of high-technology centers in the United States. Economic Development
Quarterly, 21(3): 203-222.
Spigel, B., and Harrison, R., (2017). Toward a process theory of entrepreneurial ecosystems.
Strategic Entrepreneurship Journal, 12(1): 151–168.
Stam, E., (2015). Entrepreneurial ecosystems and regional policy: A sympathetic critique. European
Planning Studies, 23(9): 1759–1769.
Stine, G., Gotsopoulos, A., and Suarez, F. F., (2015). The coevolution of technologies and categories
during industry emergence. Academy of Management Review, 40(3): 423-445.
Toole, A.A., (2012). The impact of public basic research on industrial innovation: Evidence
from the pharmaceutical industry. Research Policy, 41: 1–12.
31
Walls & Associates. (2013). “National Establishment Time-Series (NETS) Database”.
Wang, L., Madhok, A., and Xiao Li, S., (2014). Agglomeration and clustering over the industry life
cycle: Toward a dynamic model of geographic concentration. Strategic Management
Journal, 35(7): 995-1012.
Wojan, T.R., Crown, D., and Rupasingha, A., (2018). Varieties of innovation and business
survival: Does pursuit of incremental or far-ranging innovation make manufacturing
establishments more resilient? Research Policy, 47(9): 1801-1810.
Woolley, J.L., (2018). Who starts high technology firms and the relationship between
founders’ gender, educational and occupational backgrounds and firm outcomes. Academy
of Management Discoveries, online first.
Wooldridge, J. M., (2005). Simple solutions to the initial conditions problem in dynamic, nonlinear
panel data models with unobserved heterogeneity. Journal of Applied Econometrics, 20(1):
39-54.
Wu, Y., and Merriman, D., (2017). Is federal academic research and development subsidy a
complement or substitute for state funding? Public Finance and Management, 17(1): 26-47.
Zucker, L. G., Darby, M. R., and Armstrong, J., (1998). Geographically localized knowledge: spillovers
or markets? Economic Inquiry, 36(1): 65-86.
Zúniga-Vicente, J., Alonso-Borrego, C., Forcadell, F., and Galan, J., 2014. Assessing the effect
of public subsidies on firm R&D investment: A survey. Journal of Economic Surveys, 28(1):
36–67.
32
APPENDIX
Table A1: Dynamic random effects probit model with regime switching
Private Funding in Time t (Model A1)
State Funding in Time t (Model A2)
Federal Funding in Time t (Model A3)
Private funding(t-1) (Stage 1) 0.798*** -0.085 0.319** (0.156) (0.190) (0.147)
Private funding(t-1) (Stage 2) -0.708*** -0.176 -0.181 (0.200) (0.207) (0.194)
Private funding(t-1) (Stage 3) -0.699** 0.622** -0.368 (0.274) (0.265) (0.357)
State funding(t-1) (Stage 1) 0.458*** -0.070 0.349** (0.172) (0.383) (0.166)
State funding(t-1) (Stage 2) -0.323 0.072 0.169 (0.201) (0.426) (0.179)
State funding(t-1) (Stage 3) 0.722** -0.083 -0.030 (0.305) (0.442) (0.350)
Federal funding(t-1) (Stage 1) -0.091 0.058 0.743*** (0.138) (0.155) (0.210)
Federal funding(t-1) (Stage 1) -0.283 -0.050 0.040 (0.188) (0.180) (0.254)
Federal funding(t-1) (Stage 1) -0.847 0.000 -0.719** (0.524) (0.321) (0.349)
Sector 0.081*** -0.031 0.022 (0.029) (0.026) (0.029)
Total Founders 0.200* -0.112 -0.384*** (0.103) (0.090) (0.110)
Max Founder Age 0.007 -0.001 -0.001 (0.005) (0.005) (0.006)
Cumulative Education -0.006 0.020** 0.068*** (0.010) (0.010) (0.010)
Establishment Date 0.080*** 0.124*** 0.004 (0.018) (0.021) (0.016)
Industry Density -0.002*** -0.005*** -0.002*** (0.000) (0.001) (0.000)
Same Prior Industry Experience -0.110 -0.094* -0.129** (0.072) (0.057) (0.064)
Prior Entrepreneurial Experience 0.053 -0.019 0.072 (0.070) (0.061) (0.067)
Stage 2 -0.206 0.089 0.248 (0.217) (0.169) (0.231)
Stage 3 -0.518 0.376 0.294 (0.341) (0.292) (0.352)
Private funding initial 1.895*** -0.286 -0.388 (0.149) (0.228) (0.244)
Federal funding initial 0.133 -0.143 1.430***
33
(0.350) (0.331) (0.158) State funding initial -0.196 1.259*** -0.367
(0.268) (0.132) (0.239) Private funding mean 0.633** 0.668*
(0.319) (0.384) Federal funding mean 1.386*** 1.473***
(0.426) (0.325) State funding mean 1.117* 2.092***
(0.594) (0.581) Constant (Stage 1) -162.446*** -248.766*** -10.076
(36.362) (41.206) (32.532) Observations 8,057 8,057 8,057 LogL -1145.898 -565.169 -1205.356
Notes: * denotes p < 0.1, ** denotes p < 0.05, *** denotes p < 0.01. LogL is the log of the likelihood
function.
Table A2: Results of Gaussian frailty duration models for firm survival
Model A4 Model A5 Model A6 Model A7
Log (Cumulative Private Funding) -0.003 -0.026*** -0.027*** Log (Cumulative Federal Funding) -0.013** -0.015 -0.014** Log (Cumulative State Funding) 0.004 0.027 0.010 Total Founders -0.274* -0.426 -0.264 -0.264 Max Founder Age 0.012 0.019 0.012 0.011 Cumulative Education 0.008 0.008 0.006 0.007 Establishment Date 0.165*** 0.251*** 0.171*** 0.171*** Industry Density -0.004*** -0.005*** -0.004*** -0.004*** Same Prior Industry Experience -0.179 -0.388* -0.177 -0.177 Prior Entrepreneurial Experience 0.143 0.320 0.153 0.153 Log (Cumulative Number of Private Fundings) -0.110 Log (Cumulative Number of Federal Fundings) -0.313** Log (Cumulative Number of State Fundings) 0.015 Square of Log (Cumulative Private Funding) 0.000*** 0.000*** Square of Log (Cumulative Federal Funding) 0.000 Square of Log (Cumulative State Funding) -0.001 Constant -334.906*** -511.683*** -346.127*** -345.950*** lnsig2u -1.049*** 2.145*** -1.073*** -1.072*** Logarithmic Duration Dependence Yes Yes Yes Yes Observations 8,881 8,881 8,881 8,881 LogL -910.743 -902.223 -905.920 -906.128 AIC 1847.487 1830.446 1843.839 1840.256 BIC 1939.678 1922.638 1957.306 1939.539
Notes: * denotes p < 0.1, ** denotes p < 0.05, *** denotes p < 0.01. LogL is the log of the likelihood
function. AIC is the Akaike Information criteria, while BIC is the Bayesian Information Criteria.
34
Table A3: Results of gamma frailty duration models for firm survival
Model A8 ModelA9 Model A10 Model A11
Log (Cumulative Private Funding) -0.017 -0.043** -0.036*** Log (Cumulative Federal Funding) -0.043** -0.029 -0.018** Log (Cumulative State Funding) -0.011 -0.002 0.019 Total Founders -0.252 -0.251 -0.167 -0.254 Max Founder Age 0.001 0.001 0.004 0.012 Cumulative Education -0.018 -0.018 -0.017 0.005 Establishment Date 0.183** 0.180** -0.167*** 0.168*** Industry Density 0.000 0.000 -0.001 -0.004*** Same Prior Industry Experience -0.544 -0.546* -0.427 -0.215 Prior Entrepreneurial Experience -0.447 0.445 0.408 0.204 Log (Cumulative Number of Private Fundings) -0.242 Log (Cumulative Number of Federal Fundings) -0.533** Log (Cumulative Number of State Fundings) -0.109 Square of Log (Cumulative Private Funding) 0.000 0.000*** Square of Log (Cumulative Federal Funding) -0.000 Square of Log (Cumulative State Funding) -0.000 Constant -372.896** -366.507** -341.076*** -341.068 ln_varg 2.834*** 2.830*** 2.461*** 0.992***
Logarithmic Duration Dependence Yes Yes Yes Yes Observations 8,881 8,881 8,881 8,881 LogL -896.322 -896.794 -894.796 -904.444 AIC 1818.645 1819.588 1821.592 1834.888 BIC 1910.836 1911.779 1935.058 1927.079
Notes: * denotes p < 0.1, ** denotes p < 0.05, *** denotes p < 0.01. LogL is the log of the likelihood
function. AIC is the Akaike Information criteria, while BIC is the Bayesian Information Criteria.
Table A4: Detailed funding: Gaussian and gamma frailty duration models for firm survival
Model A12: Gaussian
Model A13: gamma
Received State Funding Only 0.121 0.717 Received Federal Funding Only -0.862* -1.064 Received Private Funding Only -1.727*** -2.188*** Received State & Federal Funding -1.616** -3.126** Received State & Private Funding -1.078 -2.477** Received Federal & Private Funding -4.540** -7.151*** Received State, Federal, & Private Funding -0.995 -2.045*
Reference Group: Received No Funding Total Founders -0.333 -0.256 Max Founder Age 0.018 0.003 Cumulative Education 0.013 -0.005 Establishment Date 0.239*** 0.181*** Industry Density -0.005*** -0.001
Same Prior Industry Experience -0.389** -0.537 Prior Entrepreneurial Experience 0.387* 0.667**
35
Constant -486.923*** -369.915*** lnsig2u 1.939*** ln_varg 2.737*** Logarithmic Duration Dependence Yes Yes Observations 8,881 8,881 LogL -890.902 -885.777 AIC 1815.803 1805.553 BIC 1936.362 1926.112
Notes: * denotes p < 0.1, ** denotes p < 0.05, *** denotes p < 0.01. LogL is the log of the likelihood
function. AIC is the Akaike Information criteria, while BIC is the Bayesian Information Criteria.
Table A5: Discrete Event History Analysis with Regime Model Estimation Output
Model A14:
Log (Cumulative Private Funding) (Stage 1) 0.026*** (0.007)
Log (Cumulative Federal Funding) (Stage 1) -0.041*** (0.013)
Log (Cumulative State Funding) (Stage 1) 0.034 (0.031) Stage 2 1.631*** (0.476) Stage 3 3.909*** (0.744) Log (Cumulative Private Funding) (Stage 2) -0.041*** (0.012) Log (Cumulative Private Funding) (Stage 3) -0.108*** (0.036) Log (Cumulative Federal Funding) (Stage 2) -0.020 (0.022) Log (Cumulative Federal Funding) (Stage 3) -0.097 (0.063) Log (Cumulative State Funding) (Stage 2) -0.012 (0.041) Log (Cumulative State Funding) (Stage 3) -0.121** (0.054) Total Founders -0.756***
(0.221) Max Founder Age -0.003
(0.013) Cumulative Education 0.030
(0.021) Establishment Date 0.107**
(0.043) Industry Density -0.005***
(0.001) Same Prior Industry Experience -0.402**
36
(0.161) Prior Entrepreneurial Experience 0.555*** (0.169) Log(t) 3.259*** (0.373) Constant (Stage 1) -223.442** (85.028) m2 4.744*** (0.348) logitp2 -1.413*** (0.130) Observations 8,881 LogL -873.289 AIC 1790.578 BIC 1946.595
Notes: * denotes p < 0.1, ** denotes p < 0.05, *** denotes p < 0.01. LogL is the log of the likelihood
function. AIC is the Akaike Information criteria, while BIC is the Bayesian Information Criteria.
37
TABLES & FIGURES
Table 1: Sector Classifications
Sector Definition No. Firms
Dedicated Biotechnology / Biomaterials
Basic biomaterials research, including agricultural biotech
124
Human Therapeutics Treatment of disease; drug development; clinical trials; includes some veterinary drug development
357
Services
Specialized service providers for the life sciences, including staffing and management firms, which are not involved in scientific or technological development
109
Health IT Health and life science related software and information technologies; health informatics
87
Tools / Diagnostics Biotechnology tools and diagnostics tests; use biomarkers
104
Medical Devices Instruments and machines for medical purposes 152
Table 2. Threshold regression
3 Stage Specification 4 Stage Specification No Stage Specification
(Model 1) (Model 2) (Model 3)
Stage 1 0.445*** 0.445*** 0.445***
Stage 2 0.899*** 1.467*** 0.899***
Stage 3 -0.065 0.775*** -0.065
Stage 4 -0.065
Break Dates 1997 & 2009 1997, 2004, & 2009 1997 & 2009
Notes: * denotes p < 0.1, ** denotes p < 0.05, *** denotes p < 0.01.
38
Table 3. Firms funded and amount of funding by Stage and Source
Number of Firms Funded
Stage 1:
1980-1997
Stage 2:
1998-2009
Stage 3:
2010-
Private 24 106 115
Federal 31 72 96
State 20 108 68
Total $ Amount to Firms
Private $346M $2.54B $2B
Federal $52.90M $421M $413M
State $3.78M $13.70M $18.40M
Average $ Amount to Firms
Private $14.42M $23.96M $17.39M
Federal $1.71M $3.90M $4.30M
State $0.19M $0.19M $0.27M
Table 4. Description of Covariates
Covariates Measurement Prior literature
Total Founders Number of founders of firm; proxy for early size of firm as employment unavailable
Hyytinen, Pajarinen, & Rouvinen (2015); Wojan, Crown, & Rupasingha (2018)
Max Founder Age Maximum age of any founder at time of founding
Boden & Nuccie (2000); Ortiz-Villajos & Sotoca (2018)
Founder Cumulative College
Cumulative years of college among all founders
Boden & Nuccie (2000); Dencker, Gruber, & Shah (2009); Ortiz-Villajos & Sotoca (2018)
Establishment Date Firm establishment year Kalleberg & Leicht (1991); Wojan, Crown, & Rupasingha (2018); Wooley (2018)
Industry Density
Number of active, entrepreneurial life science firms in the Research Triangle region by year, as measure of competition
Hyytinen, Pajarinen, & Rouvinen (2015); Wooley (2018)
Founder Same Prior Industry Experience
Count of founders of firm with prior work experience in life science industry
Delmar & Shane 2004; Dencker, Gruber, & Shah (2009); Hyytinen, Pajarinen, & Rouvinen (2015); Ortiz-Villajos & Sotoca (2018)
Prior Entrepreneurial Experience
Count of founders of firm with prior experience as an entrepreneur
Delmar & Shane (2004); Hyytinen, Pajarinen, & Rouvinen (2015)
Sector Categorical for all sectors except services as described in Table 1
Bruderl & Schussler (1990); Wooley (2018)
39
Table 5. Summary Statistics, all firms
All Firms
Mean Std.Dev. Min Max
Duration (years) 8.61 6.54 1 37
Log (Cumulative Private Funding) 4.11 13.65 0 159.5
Log (Cumulative National Funding) 6.91 22.70 0 364.4
Log (Cumulative State Funding) 2.24 6.47 0 66.41
Total Founders 1.83 0.99 1 10
Max Founder Age 46.91 10.35 19 74
Cumulative College 13.26 9.68 0 89
Establishment Date 1999.53 8.08 1980 2016
Industry Density 504.41 199.18 11 711
Same Industry Experience 0.92 0.90 0 9
Entrepreneurial Experience 0.54 0.81 0 4
Firm-Year Observations 10,545
Table 6. Correlations between covariates
Total Founders
Max Founder Age
Cum. College
Estab. Date
Industry Density
Same Industry Experience
Entrepren-eurial Experience
Total Founders 1.000
Max Founder Age
0.304 1.000
Cumulative College
0.852 0.323 1.000
Establishment Date
0.106 0.254 0.158 1.000
Industry Density
0.080 0.151 0.109 0.632 1.000
Same Industry Experience
0.518 0.297 0.447 0.147 0.106 1.000
Entrepreneurial Experience
0.460 0.300 0.411 0.208 0.136 0.526 1.000
40
Table 7. Dynamic random effects probit with unobserved heterogeneity estimation results
Private Funding in Time t (Model 4)
State Funding in Time t (Model 5)
Federal Funding in Time t (Model 6)
Private funding(t-1) 0.291*** -0.111 0.328** (0.108) (0.192) (0.144)
State funding(t-1) 0.411** -0.030 0.334** (0.174) (0.159) (0.164)
Federal funding(t-1) -0.084 0.041 0.708*** (0.134) (0.156) (0.115)
Sector 0.081*** -0.035 0.022 (0.028) (0.025) (0.029)
Total Founders 0.176* -0.103 -0.390*** (0.103) (0.089) (0.110)
Max Founder Age 0.008 -0.001 -0.000 (0.005) (0.005) (0.005)
Cumulative Education 0.002 -0.019* 0.068*** (0.010) (0.010) (0.010)
Establishment Date 0.081*** 0.123*** 0.005 (0.018) (0.021) (0.016)
Industry Density -0.002*** -0.005*** -0.002*** (0.000) (0.006) (0.000)
Same Prior Industry Experience -0.123* -0.088 -0.136** (0.073) (0.057) (0.064)
Prior Entrepreneurial Experience 0.054 -0.020 0.069 (0.068) (0.060) (0.067)
Stage 2 constant -0.328 0.082 0.246 (0.204) (0.162) (0.232)
Stage 3 constant -0.605* 0.480* 0.152 (0.315) (0.285) (0.350)
Private funding initial 1.855*** -0.316 -0.349 (0.145) (0.226) (0.236)
Federal funding initial 0.156 -0.119 1.396*** (0.328) (0.352) (0.149)
State funding initial -0.259 1.260*** -0.381 (0.264) (0.132) (0.237)
Private funding initial 0.833*** 0.486 (0.289) (0.329)
Federal funding mean 0.933** 1.399*** (0.407) (0.290)
State funding mean 1.391*** 2.055*** (0.507) (0.528)
Constant -165.104*** -246.499*** -12.361 (35.104) (41.095) (32.353)
Observations 8,057 8,057 8,057 LogL -1161.413 -569.192 -1210.406
Notes: * denotes p < 0.1, ** denotes p < 0.05, *** denotes p < 0.01. LogL is the log of the likelihood
function.
41
Table 8. Summary of dynamic random effects probit models with regime analysis for private
funding
Private funding in time t Stage 1 (Model 7)
Stage 2 (Model 8)
Stage 3 (Model 9)
Private funding (t-1) 0.798*** 0.081 0.099
State funding (t-1) 0.458*** 0.135 1.180***
Federal funding (t-1) -0.091 -0.374* -0.938*
Covariates Yes Yes Yes
Observations 8,057 8,057 8,057 Notes: * denotes p < 0.1, ** denotes p < 0.05, *** denotes p < 0.01.
Table 9. Summary analysis of dynamic random effects probit models with analysis for state funding
State funding in time t Stage 1 (Model 10)
Stage 2 (Model 11)
Stage 3 (Model 12)
State funding (t-1) -0.070 0.002 -0.153
Private funding (t-1) -0.085 -0.261 0.537*** Federal funding (t-1) 0.058 0.008 0.058
Covariates Yes Yes Yes Observations 8,057 8,057 8,057
Notes: * denotes p < 0.1, ** denotes p < 0.05, *** denotes p < 0.01.
Table 10. Summary analysis of dynamic random effects probit models with regime analysis for
federal funding
Federal funding in time t Stage 1 (Model 13)
Stage 2 (Model 14)
Stage 3 (Model 15)
Federal funding (t-1) 0.743*** 0.783*** 0.024
Private funding (t-1) 0.319** 0.138 -0.049
State funding (t-1) 0.349** 0.518** 0.319
Covariates Yes Yes Yes
Observations 8,057 8,057 8,057 Notes: * denotes p < 0.1, ** denotes p < 0.05, *** denotes p < 0.01.
42
Table 11. Non-parametric duration models for business survival
Non-parametric heterogeneity Model 16 Model 17 Model 18 Model 19 (preferred)
Log (Cum. Private Funding) -0.006 -0.045*** -0.044*** (0.007) (0.014) (0.013)
Log (Cum. Federal Funding) -0.028** -0.040*** -0.034*** (0.011) (0.015) (0.011)
Log (Cum. State Funding) -0.017 0.017 -0.008 (0.023) (0.046) (0.021)
Total Founders -0.770*** -0.879*** -0.759*** -0.760*** (0.271) (0.297) (0.255) (0.255)
Max Founder Age 0.002 0.024 0.000 0.003 (0.016) (0.016) (0.014) (0.013)
Cumulative Education 0.033 0.050* 0.029 0.030 (0.022) (0.027) (0.022) (0.023)
Establishment Date 0.160*** 0.161*** 0.215*** 0.213*** (0.056) (0.000) (0.044) (0.042)
Industry Density -0.002 -0.002*** -0.004** -0.004** (0.002) (0.001) (0.002) (0.002)
Same Prior Industry Experience -0.353* -0.417** -0.289* -0.297* (0.181) (0.166) (0.164) (0.164)
Prior Entrepreneurial Experience 0.362* 0.273 0.277 2.909*** (0.186) (0.172) (0.175) (0.171)
Log (Cum. Number of Private Fundings) -0.126 (0.094)
Log (Cum. Number of Federal Fundings) -0.318** (0.124)
Log (Cum. Number of State Fundings) -0.355 (0.256)
Square of Log (Cum. Private Funding) 0.000*** 0.000*** (0.000) (0.000)
Square of Log (Cum. Federal Funding) 0.000 (0.000)
Square of Log (Cum. State Funding) -0.000 (0.002)
Constant -328.463*** -330.409 -438.385*** -434.837*** (112.470) (.) (88.494) (84.480)
m2 4.548*** 4.688*** 4.683*** 4.666*** (0.444) (0.381) (0.368) (0.359)
logitp2 -1.585*** -1.601*** -1.591*** -1.596*** (0.146) (0.133) (0.135) (0.135)
Logarithmic Duration Dependence Yes Yes Yes Yes
Observations 8,881 8,881 8,881 8,881 LogL -896.343 -898.025 -889.722 -890.028 AIC 1820.686 1822.049 1813.445 1810.056 BIC 1919.969 1914.241 1934.003 1916.431
Notes: * denotes p < 0.1, ** denotes p < 0.05, *** denotes p < 0.01. LogL is the log of the likelihood
function. AIC is the Akaike Information criteria, while BIC is the Bayesian Information Criteria.
43
Table 12. Duration models for business survival by funding source(s)
Non-parametric heterogeneity Model 20
Received State Funding Only 0.191 (0.471)
Received Federal Funding Only -0.929 (0.623)
Received Private Funding Only -1.319** (0.537)
Received State & Federal Funding -2.222** (1.081)
Received State & Private Funding -1.638** (0.638)
Received Federal & Private Funding -4.756*** (1.132)
Received State, Federal, & Private Funding -1.915*** (0.580)
Reference Group: Received No Funding Total Founders -0.730***
(0.278) Max Founder Age -0.005
(0.015) Cumulative Education 0.041
(0.026) Establishment Date 0.165***
(0.062) Industry Density -0.003
(0.002) Same Prior Industry Experience -0.310*
(0.183) Prior Entrepreneurial Experience 0.408**
(0.199) Constant -336.727***
(123.491) m2 4.554***
(0.371) logitp2 -1.445***
(0.170) Logarithmic Duration Dependence Yes
Observations 8,881 LogL -884.927 AIC 1805.854 BIC 1933.504
Notes: * denotes p < 0.1, ** denotes p < 0.05, *** denotes p < 0.01. LogL is the log of the likelihood
function. AIC is the Akaike Information criteria, while BIC is the Bayesian Information Criteria.
44
Table 13. Summary analysis of discrete event history models with regime analysis
Stage 1 (Model 21)
Stage 2 (Model 22)
Stage 3 (Model 23)
Log (Cum. $ Private) 0.026*** -0.015 -0.082***
Log (Cum. $ Federal) -0.041*** -0.061*** -0.138**
Log (Cum. $ State) 0.034 0.022 -0.087*
Logarithmic Duration Dependence Yes Yes Yes
Covariates Yes Yes Yes Observations 8,881 8,881 8,881
Notes: * denotes p < 0.1, ** denotes p < 0.05, *** denotes p < 0.01.
45
Figure 1. Annual No. of Active RT Region Entrepreneurial Life Science Firms
Source: PLACE: RTP Database.
Figure 2. Annual No. of RT Entrepreneurial Life Science Firms Receiving Funding by Source
Source: PLACE: RTP, adapted from CB Insights, NIH Reporter, SBA, NC IDEA, NC Biotech Center.
0
100
200
300
400
500
600
7001
98
0
19
81
19
82
19
83
19
84
19
85
19
86
19
87
19
88
19
89
19
90
19
91
19
92
19
93
19
94
19
95
19
96
19
97
19
98
19
99
20
00
20
01
20
02
20
03
20
04
20
05
20
06
20
07
20
08
20
09
20
10
20
11
20
12
20
13
20
14
20
15
20
16
0
10
20
30
40
50
60
70
80
90
100
19
80
19
81
19
82
19
83
19
84
19
85
19
86
19
87
19
88
19
89
19
90
19
91
19
92
19
93
19
94
19
95
19
96
19
97
19
98
19
99
20
00
20
01
20
02
20
03
20
04
20
05
20
06
20
07
20
08
20
09
20
10
20
11
20
12
20
13
20
14
20
15
20
16
Private Federal State Total
46
Figure 3. Annual Real Dollar Amount of Funding to RT Entrepreneurial Life Science Firms by Source
Source: PLACE: RTP, adapted from CB Insights, NIH Reporter, SBA, and NC Biotech Center.
0
50
100
150
200
250
300
350
400
450
500
550
600
650
19
80
19
81
19
82
19
83
19
84
19
85
19
86
19
87
19
88
19
89
19
90
19
91
19
92
19
93
19
94
19
95
19
96
19
97
19
98
19
99
20
00
20
01
20
02
20
03
20
04
20
05
20
06
20
07
20
08
20
09
20
10
20
11
20
12
20
13
20
14
20
15
20
16
Mill
ion
s
Private State Federal
Figure 4. RT Region Life Science Industry Timeline.
Source: Information taken from Smilor et al. (2007), Hardin (2008), Feldman & Lowe (2011), Lowe & Feldman (2015), McCorkle (2012), a
variety of online sources, and interviews with local entrepreneurs and ecosystem members.
Note: Green font indicates a funding relate event. Red font indicates a university related event. Blue font indicates public event.
• NC Biotech Center, 1981
• Quintiles founded by UNC prof, 1981
• Glaxo HQ move, 1983 • Council for Entrepreneurial Development, 1983
• NC State TTO, 1984
• Syngenta move, 1984
• Intersouth Partners founded as first local VC firm, 1985
• PPD founded by UNC prof, 1985
• Sphinx Pharmaceuticals, 1986
• Duke TTO, 1986
• BASF Corp. Agricultural Products, 1986
• NC Biotech Center offers first business loans, 1989
• First Flight Venture Center, 1991 • Embrex IPO ($16.7M), 1991
• Sphinx IPO ($75M), 1991
• Biotechnology Industry Organization (BIO), 1993
• LabCorp move, 1993
• Sphinx sold to Eli Lilly, 1994 • The Aurora Funds and Pappas Capital (VC), 1994
• UNC TTO, 1995 • Burroughs Wellcome & Glaxo merger, 1995, freed
talent
• Fujifilm Diosynth (CMO) move, 1995
• Aeolus Pharmaceuticals & Quintiles IPOs, 1996
• Mycosearch sold to Onogene ($4.7M)
• Significant expansion and turnover in ESOs
• Quintiles IPO-II ($947M) & LipoScience IPO ($45M), 2013
• Advanced Liquid Logic sold to Illumina ($96M), 2013
• KBI BioPharma (CMO) move, 2013
• Humacyte raises $150M Series B, 2015
• Bamboo Therapeutics sold to Pfizer ($150-645M), 2016
• 10+ other IPOs including Liquidia ($50M) and G1 Therapeutics ($105M)
• University Angel Networks: CAN, DAN, WIN
• BioLabs NC, 2018
• Firms successful in getting nonlocal VC
• NC ranked #4 in biotech VC investment, 1998
• Eno River Capital (VC), 1998
• BD BioVenture Center, 1999
• Icagen raises $19M in VC, 1999
• Inspire Pharmaceuticals IPO ($69M), 2000
• Glaxo-Wellcome & SmithKline Beecham merger, 2000
• Hatteras Venture Partners (VC), 2000
• NC State Centennial Biomedical Campus, 2000
• Triangle Pharmaceuticals raises $75M in VC, 2001
• Trimeris IPO ($102M), 2002
• Southeast TechInventures, 2003
• Icagen IPO, 2005
• One NC SBIR Match Fund, 2006-2011, 2016-present
• Dara Biosciences IPO, 2008
• Hamner Bioscience Accelerator, 2008
• Cempra Pharmaceuticals raises $50M Series F, 2009
1980 - 1997 1998 - 2009 2010 - Present
DOCUMENTS DE TRAVAIL GREDEG PARUS EN 2019GREDEG Working Papers Released in 2019
2019-01 Muriel Dal Pont Legrand & Harald Hagemann Impulses and Propagation Mechanisms in Equilibrium Business Cycles Theories: From Interwar Debates to DSGE “Consensus”2019-02 Claire Baldin & Ludovic Ragni Note sur quelques limites de la méthodologie de Pareto et ses interprétations2019-03 Claire Baldin & Ludovic Ragni La conception de l’homme dans la théorie de l’Echange Composite de François Perroux : entre homo economicus et homo religiosus2019-04 Charlie Joyez Shared Ownership in the International Make or Buy Dilemma2019-05 Charlie Joyez Alignment of Multinational Firms along Global Value Chains: A Network-based Perspective2019-06 Richard Arena & Ludovic Ragni Nature humaine et choix rationnel : Pareto contre Walras ?2019-07 Alain Raybaut A First French Episode in the Renewal of Nonlinear Theory of Economic Cycles (1978-1985)2019-08 Alain Raybaut Bertrand Nogaro et l’économie de guerre : le Service de la main d’œuvre étrangère2019-09 Nicolas Brisset & Dorian Jullien Models as Speech Acts: A Restatement and a new Case Study2019-10 Kozo Kiyota, Toshiyuki Matsuura & Lionel Nesta On Export Premia2019-11 Nicolas Brisset & Raphaël Fèvre Peregrinations of an Economist: Perroux’s Grand Tour of Fascist Europe2019-12 Marco Baudino Urbanization and Development: A Spatial Framework of Rural-to-urban Migration2019-13 Giuseppe Attanasi, Kene Boun My, Nikolaos Georgantzís & Miguel Ginés Strategic Ethics: Altruism without the Other-regarding Confound2019-14 Thierry Kirat & Frédéric Marty How Law and Economics Was Marketed in a Hostile World: L’institutionnalisation du champ aux États-Unis de l’immédiat après-guerre aux années Reagan2019-15 Agnès Festré, Ankinée Kirakozian & Mira Toumi La menace est plus forte que l’exécution, mais pas pour tous : sanction versus recommandation par un tiers dans une expérience de bien public2019-16 Nicolas Brisset, Raphaël Fèvre & Tom Juille Les années noires de la “Science de l’Homme”: François Perroux, la Fondation Carrel et l’appropriation de la sociologie2019-17 Romain Plassard From Disequilibrium to Equilibrium Macroeconomics: Barro and Grossman’s Trade-off between Rigor and Realism2019-18 Christophe Charlier, Gilles Guerassimoff, Ankinée Kirakozian & Sandrine Selosse Under Pressure! Nudging Electricity Consumption within Firms: Feedback from a Field Experiment
2019-19 Nicolas Brisset & Raphaël Fèvre Prendre la parole sous l’État français: le cas de François Perroux2019-20 Giuseppe Attanasi, Ylenia Curci, Patrick Llerena & Giulia Urso Intrinsic vs. Extrinsic Motivators on Creative Collaboration: The Effect of Sharing Rewards2019-21 Giuseppe Attanasi, Ylenia Curci, Patrick Llerena, Adriana Carolina Pinate & Giulia Urso Looking at Creativity from East to West: Risk Taking and Intrinsic Motivation in Socially and Culturally Diverse Countries2019-22 Nobuyuki Hanaki Cognitive Ability and Observed Behavior in Laboratory Experiments: Implications for Macroeconomic Theory2019-23 Dongshuang Hou, Aymeric Lardon, Panfei Sun & Genjiu Xu Sharing a Polluted River under Waste Flow Control2019-24 Sothearath Seang & Dominique Torre Proof of Work and Proof of Stake Consensus Protocols: A Blockchain Application for Local Complementary Currencies2019-25 Paige Clayton, Maryann Feldman & Benjamin Montmartin Funding Emerging Ecosystems